Correct, I was not saying to keep your valves open as long as possible, I was saying maximise the fully open time for a given phase requirement (i.e. you want to open and close the valves to fully open/closed as quickly as possible and hold them there for as much of your open duration as you can). An ideal valve (as far as minimising pumping losses goes) would fully open instantaneously, hold open, and then close instantaneously (I think this is what F1 were trying to achieve with their pneumatic systems). I think this is true of any engine (a valve on its way up or down is simply adding friction (mechanical and aerodynamic) to the equation). However, with our restricted engines, a higher maximum lift cam won't necessarily work (as you have said), which is why some teams regrind for smaller lift (increasing the aerodynamic friction to create swirl, thus improving combustion efficiency even if the pumping efficiency side is reduced).
Anyway, to work out the safe mm/degree of your theoretical cam, you need to follow the advice I gave above regarding valve bounce. Essentially you need to work out the mass of the valve assembly, how fast it is going to move due to a given cam profile at a given maximum RPM, how much the spring is going to resist, and whether or not the momentum caused in the valve assembly in this system is enough to overcome the force that the spring applies to keep the valve/follower in contact with the cam lobe. Given that you appear to be interested in reducing lift you probably won't have to worry about valve bounce. So I would take tromoly's advice and analyse the stock profile (using the experimental method I outlined above), which will produce a pretty picture. You can then essentially scale down that picture to create your cam profile. Just make sure you don't go crazy with ramp angles (i.e. KISS).